Blog Benchmarks · After Andromeda
Meta got better at the click and worse at the lead
We compared $3.1 million of Meta spend for 48 membership businesses, a year before Andromeda and a year after. The dashboard improved. The cost of a lead went up.
If you own a gym, a studio, a med spa or a wellness clinic, you may have had an odd year with Meta. The reports look healthy. More people click. Each click costs less. And yet the leads feel more expensive, and the front desk is not busier.
We had the same feeling in the accounts we run, so we measured it. Andromeda is the system Meta now uses to decide which ad each person sees, and it changed how the auction rewards creative for every advertiser at once. We wanted to know what that change did to businesses that sell memberships, using the numbers from our own accounts.
The short answer: Meta got better at the click and worse at the lead.
Source: After Andromeda, Sneeze It, 46 businesses, $3.06M of spend
Every click number improved. The lead got more expensive.
Those four numbers come from 46 of the 48 businesses in the study, the ones with at least 20 leads in each window, covering $3.06 million of spend. The first window runs from October 2024 to September 2025. The second runs from October 2025 to September 2026. The same businesses sit on both sides, so this is a business compared with itself a year later.
Read the first two rows and you would call it a good year. Those are the numbers most dashboards put at the top, and the ones many monthly reports lead with.
The third row is where the year turned. Of the people who clicked, fewer filled in the form: 8.8% the year before, 6.9% the year after. That is a 22% drop, and it was large enough to swallow the cheaper clicks. Cost per lead went from $28.86 to $31.70.
The arithmetic is simple, and it is worth keeping in your head. Cost per lead is the cost of a click divided by the share of clicks that turn into leads. If clicks get a little cheaper and conversion falls further, every lead costs more, even though nothing on the click side of the report looks wrong.
A cheaper click is only good news if the same share of people still walk through the door.
Why would this happen? The study does not claim to know what is inside Meta's system. It shows the pattern. Meta became very good at finding people who will tap an ad. Tapping is easy. Giving a name and a phone number to a gym is a bigger step, and the people Meta now finds for the tap are less likely to take it.
What the dashboard shows
More clicks, a higher click-through rate and a lower cost per click. Month over month, the click-based numbers trend the right way.
What the business feels
Fewer of those clicks become leads, so each lead costs more. The spend goes up to keep the same number of people coming in.
What else is inside the study
Finding 01 is the hinge, and the other four findings are about creative: which kinds of ads held up after the change. A pattern only counts in the study when it beat the rest of its own account in most accounts, so one big advertiser cannot carry a result.
Finding 02, format. Video wins the click. Static wins the lead. Inside the same accounts, video pulled more clicks and still cost more per lead in most of them. A static ad was cheaper per lead in 62% of accounts. This fits finding 01 closely: the format that is best at earning a tap is not the format that is best at earning a form.
Finding 03, copy. One copy choice flipped after Andromeda. It lost more often than it won the year before. The year after, it won in 11 of 14 accounts.
Finding 04, offer. Two offers went from losing to winning. Each one lowers the cost of the first step, and each now brings leads at a median of 17% less than the rest of its account.
Finding 05, held both years. There is a line owners love to put in an ad. It lost in roughly three of four accounts, before the change and after, with a median cost per lead 16% higher than the rest of the account.
The specifics of 03, 04 and 05 stay in the study, with a chart and an action item for each finding and a one-page build list for whoever makes your ads. If you want the background on the update, we wrote about what Meta's Andromeda update means for businesses buying ads in 2026, and about how we grade creative in what 18,920 ads taught us about running yours.
How we did it, and what it cannot tell you
We took 48 membership businesses that spent on Meta in both 12-month windows: gyms, studios, med spas, aquatics, and wellness and IV clinics. Together that is $3.1 million of spend, 100,786 leads and 14,549 ads.
For the creative findings, every ad was compared to the rest of its own account. A static ad was never compared to a video from a different business in a different town, only to the other ads its own business ran in the same window. A pattern only counts in an account where both sides spent at least $500 and brought in at least 5 leads, so a handful of lucky leads cannot decide anything.
The study has limits. This is not a controlled experiment. Nobody randomly assigned ads to formats or offers, and plenty changed in those businesses over two years besides Meta's system. The before and after numbers in finding 01 are a real description of what happened in these accounts. The creative findings are directions to test in your own account, not laws.
What to check in your own account this week
Your own numbers from a year apart will show whether this happened to you.
- Put cost per lead at the top of the report
Click-through rate and cost per click can both improve while each lead gets more expensive. Judge the account on what a lead costs.
- Compare to your own account a year ago
Pull October 2024 to September 2025 next to the last 12 months. Your own history is the fairest benchmark you have.
- Check how many clicks become leads
Divide leads by link clicks for both years. If that share fell, it explains most of what you are feeling.
- Split it by format
Look at cost per lead for static and video ads separately inside your account, not across the industry.
- Revisit your target
A cost per lead goal set in 2024 was set in a different auction. Decide whether it still describes the market you are buying in.
The last step is the uncomfortable one. Many owners still hold their ads to a number from before the change, and then judge every month as a failure against it. The study has a section on why a 2024 cost per lead is the wrong yardstick now, and how to set a new one.
Sources
- Sneeze It, After Andromeda: More Clicks, Fewer Leads, November 1, 2026, https://benchmark.sneeze.it/andromeda/
